ABAQUS Austria Users Conference

November 24-25, 2003

GRAND HOTEL Wien
Krntner Ring 9
A-1010 Vienna

FE - Simulation of Thin Strip and Temper Rolling Processes

Alexander Kainz, Konrad Krimpelsttter, Klaus Zeman
Institute for Robot & Manufacturing Systems
Department for Computer Aided Methods in Mechanical Engineering
Johannes Kepler University of Linz, Altenbergerstr. 69, A-4040 Linz, Austria.

Key Words: Non-Linear Finite Element Analysis, Contact and Friction, Large Deformations, Elasto-plasticity,
Work Roll Flattening, Temper Rolling, Contained Plastic Flow.

Abstract
Within the production chain of cold-rolled flat products skin-pass and temper rolling represent the
final forming stage, in which material and surface properties as well as flatness are tailored to satisfy
even the most demanding customer requirements. Almost all skin-pass and temper-mill setups are
based on offline calculations, supplemented by trial and error procedures during operation. To
minimize strip segments that do not match the prescribed tolerances, which occurs especially after
changes in sheet sizes, it is of vital importance to incorporate appropriate offline and online models to
the automation structure of such rolling mills. Key objectives of adequate models are the precise
determination of the rolling force, of the roughness transfer and of the roll-flattening behaviour. As is
well known for thin sheet, foil and temper rolling processes, the simplified assumption of a circular arc
of contact breaks down. Therefore, the correct non-circular surface contour of the work roll has to be
determined simultaneously with the resulting contact pressure distribution between strip and work roll.
To attain a comprehensive understanding of these underlying process details, and to check and tune
semi-analytical model approaches (cf. e.g. [7-10]), highly sophisticated numerical approaches based
on the method of finite elements have been performed by utilizing the non-linear capabilities of
Abaqus Standard and Explicit [1]. Due to extremely short contact lengths in temper rolling processes,
a very fine spatial discretization is essential, often leading to several hundred thousand degrees of
freedom even for plane strain calculations. Special emphasis has to be put on work roll flattening [6]
and on the formation of contained plastic flow regions [2].

Introduction and Survey

Rolling processes can be considered to be key steps within the production of steel. Therefore,
the development of highly sophisticated mathematical process models is a vital precondition for
manufacturing high quality products satisfying narrow tolerance demands. By performing systematic
parameter studies and regression methods, sets of characteristic curves can be obtained, which serve as
an input for online control and process automation systems. Mathematical online and offline models,
validated against each other and calibrated with real process data, are essential to determine proper

1/8

Abaqus Austria Users Conference 2003, Vienna, Austria

mill setups, to improve the quality of the rolled product, to control the online process of metal forming,
and to optimize the mill throughput and yield (i.e. minimize scrap).
In strip mill operations, the temper rolling of steel represents the final stage within the production chain of cold-rolled flat products. With annealed strips to be subsequently formed, temper rolling
provides a slight reduction in thickness, and thereby eliminates the yield point elongation in the stressstrain curve of the steel. This permits the material to be formed without developing Luders bands.
Besides, temper rolling is also used to improve the surface quality of the strip and the flatness
properties. To optimize these final properties of cold rolled steel, the elongation of each strip product
must be strictly controlled to a certain value in the temper rolling process. Therefore, an accurate
elongation control system, which is based on precise mathematical process models for the prediction
of roll force, torque and forward slip is essential. The key criteria in skin-passing are surface
cleanliness and roughness, which are important for subsequent painting, welding and deformation
processes. An important issue is to keep the surface of the work rolls clean to prevent that particles are
imprinted into the strip in the roll gap. This task can be accomplished by utilizing either the
conventional wet temper system, where a water-agent mixture is sprayed into the entry of the roll bite,
or by the dry temper technique (Fig. 1), where a combination of rotating and oscillating brushes is
pressed slightly against the roll surfaces thereby cleaning all particles sticking to the roll surfaces [7].
Backup-Roll Brushes
Backup Roll

Adjustment Cylinders
Flatness Measurement Roll

Work Roll

Work-Roll Brushes Bridles

Fig. 2: Decomposition of the deformation zone

inside the strip in elastic and plastic regions

Fig. 1: State of the Art Dry Temper Rolling

Equipment (cf. [7])

For temper rolling processes the conditions inside the roll bite are significantly different from those of
usual hot or cold rolling processes in reversing or tandem mills. The work rolls are significantly
deformed in a way similar to that of Hertzian contact. The strip elongation (or reduction) is of the
order of one percent (~0.2%2%), the deformation is highly localized and a significant elastic springback of the strip can be observed. Therefore, standard circular arc roll gap models (e.g. Bland Ford
Ellis combined with an effective Hitchcock curvature) do not apply. Such models predict huge
pressure increases in the roll gap when rolling thin gauges due to the frictional forces present
throughout the roll gap, combined with the simplifying assumption of a circular arc of contact. A
significant improvement was achieved by Jortner et al. [6] describing the work roll deformation in
radial direction using influence functions. In 1987 Fleck and Johnson [2] published a theory
appropriate for the rolling of thin strips and foil. Their theory omits the use of simplifying
presumptions on roll-flattening geometry, which can be considered as main reason for failure when
applied to thin-gauge rolling. Besides, their model predicts the occurrence of an intermediate
contained plastic flow zone, where a flat work roll surface parallel to the strip middle plane occurs.
Because of the trend to thinner and harder strips, and due to the fact that the contact length is
very short for small strip elongations, the development of improved offline and online models for
temper rolling is essential. Enhanced models developed recently [3, 7-10] include an elastic
compression zone at the roll gap entry, an elastic recovery zone at the exit and a plastic zone in

2/8

Abaqus Austria Users Conference 2003, Vienna, Austria

between, whereby elastic regions are also allowed to arise between plastic zones (Fig. 2). This ensures
that a contained plastic flow zone is included automatically in the model and need not be postulated a
priori. To ensure praxis-relevant results a systematic comparison and calibration of these highly
sophisticated semi-analytical models with real process data and with results obtained by Finite element
simulations is absolutely necessary. Transient dynamic Finite Element simulations, which are of
course much more calculation time consuming than problem-adequate optimized steady-state 1D
solution procedures (nevertheless highly non-linear and ill posed problems), enable a deep insight into
the underlying process details and yield valuable feedback to the qualitative solution behaviour.

Method of Analysis

To obtain a deeper understanding of the occurrence of contained plastic flow regions inside
the roll gap for temper rolling processes with small strip elongations, we decided to perform a transient
dynamic simulation of this process by utilizing the non-linear capabilities of the software package
Abaqus Explicit [1]. Due to the geometric properties of the setup (strip width of 1288 mm), a plane
strain 2D simulation of the vertical mid-plane suffices. Note that on account of a horizontal mid-plane
symmetry property only the upper work roll and the upper half of the strip have to be simulated. A
strip of initial thickness of 500 m is rolled to an elongation of 0.5% by an elastically deformable
work roll with undeformed radius of 295 mm and prescribed angular velocity of 22.63 rad/s. The strip
is modeled as an elasto-plastic 2D solid body. It should be emphasised that a strip reduction of 0.5%,
i.e. the upper half of the strip is reduced from 250 m to 248.7 m, is accomplished by lowering the
center of the work roll by 50 m, i.e. the vertical work roll shift exceeds the strip draft by a factor 40.
At least 5 elements in (half) thickness of the strip are necessary to obtain reliable stress and strain
distributions inside the strip. Back and front tension stresses of 25 and 43 MPa, respectively, are
applied to the inlet and outlet cross section of the strip and are kept fixed during the whole simulation.
The initial velocity of the strip is put to 400 m/s, which coincides with the initial circumferential
velocity of the work roll surface and ensures minimum possible strip element distortions at the initial
phase of the simulation, when the strip is pulled into the roll gap. The yield stress curve includes strain
hardening. For zero logarithmic strain the yield stress is assumed to be 400 MPa, increasing to
440 MPa for 1% strain, 500 MPa for 2.5% and up to 800 MPa for an equivalent strain value of 10%.
For the first simulation-runs a strain rate dependence was neglected but will be incorporated in future
investigations. The Youngs modulus of the strip is chosen as 185 GPa and 225 GPa for the work roll.
The Poissons ratio for steel is =0.3 and the mass density takes the value =7.85 kg/dm3. The friction
between the strip (slave) surface and the work roll (master) surface is modeled by utilizing the
Coulomb friction law with a friction value of =0.15. In Abaqus Explicit the kinematic surface
interaction algorithm with hard overclosure (describing the normal contact behaviour) turned out to be
significantly superior to the penalty formalism, where an inevitable penetration of the master work roll
surface into the strip lead to an extreme distortion of the strip contact surface elements.
Due to the very short contact lengths an extremely fine discretization of the work roll sleeve
near the contact region is essential to obtain reliable and praxis-relevant data. In order to calculate at
least 20 stress and displacement values along the non-circular arc of contact, a value of 200 m or
even shorter has to be chosen for the spatial discretization length of the work roll master surface. As
the corresponding strip slave surface element length has to be chosen shorter, i.e. about 50-100 m,
the stable time increment for explicit time integration turns out to be of the order of 10-9 sec.
(determined by Abaqus Explicit without mass-scaling) and for implicit methods of about 10-7 sec.
Numerical considerations pointed out that at least 5-10 contact lengths have to be rolled, until a well
pronounced steady state is reached. This results in a real process time of at least 4-8 ms for the test
case under consideration. Therefore, the utilization of implicit solution concepts (Abaqus Standard) is
not appropriate due to the extremely large number of intermediary linear stiffness equations that have
to be solved, whereas explicit dynamic algorithms (Abaqus Explicit) are much better suited for this
class of problems. Moreover, the memory requirements in Abaqus Explicit are one order of magnitude

3/8

Abaqus Austria Users Conference 2003, Vienna, Austria

less compared to the Standard simulation variant. To speed up the calculation, mass scaling was
employed and will be discussed in more detail in the next section. Special emphasis has to be put on an
adequate modeling of the elastically deformable work roll (Fig.3). The inner ring of the work roll is
modeled as rigid body driven by an angular velocity constraint condition imposed on the reference
point in the work roll centre. Previously performed systematic studies [4, 5] pointed out that such a
simplification does not adulterate the deformation behaviour of the work roll sleeve as long as the
radius of the inner rigid body region remains small compared to the outer work roll radius. On account
of the very fine discretization of the work roll sleeve, the overall number of degrees of freedom would
become very large for the case of an axially symmetric mesh, leading to a system with several 100 000
degrees of freedom. Therefore, only a 30 sector of the work roll sleeve was meshed very fine,
resulting in a dramatic reduction of the number of degrees of freedoms (a total of 119 000 variables for
the whole model incl. strip) without significant loss of numerical accuracy. The underlying work roll
partition concept is represented in Fig. 4. The transition from a coarse structured inner mesh to ever
finer outer meshes is performed by using concentric layers of triangular elements such that the mesh
compatibility condition is never violated, as can be seen in Fig. 5. A detailed view of the work roll
sleeve and the transition to coarser mesh regions in both circumferential and radial directions is given
in Fig.6.

Fig. 3: Combined structured (quadrilateral) and free

(triangular) meshing concept for the elastically
deformable work roll

Fig. 4: The underlying work roll partitions to reduce

the overall number of degrees of freedom without
significant loss of accuracy

Fig. 5: Transition between coarse and fine meshed

regions by triangular elements

Fig. 6: Detailed view of the relevant fine discretized

work roll sleeve sector

4/8

Abaqus Austria Users Conference 2003, Vienna, Austria

Numerical Results

The temper rolling process under consideration, as specified in the preceding section, was
simulated by utilizing Abaqus Explicit. To obtain reliable steady state results, at least 5-10 deformed
contact lengths have to be simulated. For this test case with a strip velocity of about 6.67 mm/ms, the
rolling of a strip length of 150 mm takes approximately 22 ms real process time. This requires about
~106-107 time increments, as the stable time increment is of the order of ~10-8-10-9 s (estimated by
Abaqus Explicit), leading to a simulation calculation time of several days. To speed up the calculation
mass scaling turned out to be very attractive. However, results obtained by choosing high mass scaling
factors have to be analyzed with due care and lead to fast wrong answers, as is obvious from Table 1.
Mass scaling
factor
[dim.less]

Number of
increments
[dim. less]

Mises stress
Max. in roll
[Mpa]

Mises stress
Max. in strip
[Mpa]

Gap entry
position [mm]

Gap exit
position
[mm]

Proj. Gap
Length [mm]

Strip exit
thickness
[m]

509262

498

444

-2.46

2.27

4.72

247.83

10

161040

509

435

-2.45

2.35

4.80

248.05

100

50923

530

427

-2.54

2.32

4.86

248.58

1000

16584

1012

562

-4.56

3.89

8.45

243.79

10000

5431

2565

800

-9.97

8.22

18.20

230.11

Table 1: Comparison of some relevant process data, obtained by different mass scaling values and evaluated at
the process time 2 ms, when a strip length of 13.35 mm is rolled.

For this transient dynamic simulation, where an elastically deformable work roll and an elasto-plastic
strip (and of course contact modeling in between) are involved, a fixed mass scaling factor of 100,
applied to the whole model at the beginning of the simulation, proved to be a satisfactory compromise
between accuracy and calculation time, where the latter quantity is reduced by a factor of 10 compared
to no mass scaling.

Fig. 7: Mises stress distribution inside the strip and

work roll after 0.1 ms of process time

Fig. 8: The same distribution after 1 ms process time

Further advanced simulation techniques available in Abaqus Explicit include adaptive meshing on
either Lagrangian or Eulerian domains, which turned out to be very beneficial for those test runs,
where significantly higher strip reductions were prescribed. One of the most critical aspects is the
accurate determination, whether a steady state has already been reached. For the class of rolling
simulations treated in this context it turned out that a reliable criterion for steady state detection is the
ratio between the increase of the kinetic energy per unit time and the dissipated power. Provided that
this ratio is small compared to unity a steady state scenario can be assumed.
Of particular interest in dynamic transient FE simulations is the development of stress and deformation
fields during the time evolution. The Mises stress distributions inside the strip and the relevant part of
the work roll are represented in Fig. 7 and Fig. 8 after 0.1 ms and 1 ms real process time, respectively.

5/8

Abaqus Austria Users Conference 2003, Vienna, Austria

The stationary Mises stress distribution after 20 ms process time, when already 27 contact lengths are
rolled, is depicted in Fig. 9. An enlarged view of the relevant part inside the work roll is given in
Fig. 10. These results meet the expectations and justify the work roll modeling concepts outlined in the
preceding section.

Fig. 9: Stationary Mises stress distribution after 20

ms of real process time

Fig. 10: Detailed view of the steady state Mises

distribution inside the work roll

This FE-simulation yields valuable information about the emerging stress-, velocity- and displacement
fields inside the work roll and the roll gap. Moreover, the non-circular arc contour of the work roll is
determined simultaneously with the corresponding contact pressure and shear stress distributions. A
typical pressure profile for steady-state dry temper rolling is shown in Fig. 11, together with the
corresponding shear stress distribution in Fig. 12 for the case of Coulomb friction with a friction
coefficient =0.15.

Fig. 11: Dry temper rolling steady-state contact

pressure distribution along the arc of contact

Fig. 12: The corresponding shear stress distribution

for the case of Coulomb friction with =0.15

The same contact and shear stress distributions can be obtained by an evaluation along the path of a
selected node at the strip surface when moving through the roll gap. These results are depicted in
Fig. 13 and agree extremely well with the ones taken at one point in time as represented in Fig. 11 and
Fig. 12. This coincidence fully confirms that a steady state has been reached. The stationary work roll
surface contour is shown in Fig. 14, and the corresponding vertical and horizontal strip surface
displacement fields (relative Lagrangian representation) are depicted in Fig. 15 and Fig. 16,
respectively. A well pronounced contained plastic flow zone inside the roll gap can be identified,
where the contact surface contour is largely parallel to the abscissa and no strip reduction occurs.
Inside this region the slip rate FSLIPR1, i.e. the slip velocity component in tangent direction, remains
(approximately) zero along the contact line, as shown in Fig. 18, and the accumulated slip
displacement FSLIP1 is almost constant inside this region, as is represented in Fig. 17.

6/8

Abaqus Austria Users Conference 2003, Vienna, Austria

Fig. 14: The deformed steady state contact surface

contour between the work roll and the strip inside
the roll gap

Fig. 13: Contact pressure and corresponding shear

stress values acting on a selected node while
moving through the roll gap

Fig. 15: The vertical strip displacement component U2

for the steady case inside the roll gap

Fig. 16: The corresponding relative Lagrangian horizontal displacement component U1 of the strip surface
inside the roll gap

Fig. 18: The corresponding contact slip rate in

longitudinal direction FSLIPR1 along the arc of
contact for the steady state case

Fig. 17: The accumulated slip displacement

component FSLIP1 along the contact surface for the
steady-state case

7/8

Abaqus Austria Users Conference 2003, Vienna, Austria

Conclusions and Future Prospects

In the present study a typical temper rolling process was simulated by utilizing the finite element
package Abaqus Explicit [1]. Such investigations yield valuable information about the stress-,
velocity- and displacement fields inside the roll gap. Moreover, the non-circular arc contour [5-7] of
the work roll is determined simultaneously with the corresponding contact pressure and shear stress
distributions, and a contained plastic flow zone [2-4, 7] inside the roll gap is identified. Due to the
very short contact lengths an extremely fine discretization of the work roll sleeve near the contact
region is essential to obtain reliable and praxis-relevant data. Further research objectives comprise
systematic parameter studies of the temper rolling process and a critical analysis of the results
calculated by FE-simulations. The data obtained by using Abaqus Explicit will be validated against
real process data and compared with results determined by highly sophisticated semi-analytical models
based on an extended version of the Jortner Greens function approach [5, 6] for the elastic work roll
deformation, combined with an enhanced version of the Fleck Johnson [2, 7-10] strip model including
elastic compression, recovery and contained plastic flow regions.

Acknowledgements:
The company HKS (Hibbitt, Karlsson & Sorensen Inc.) is acknowledged for the provision of the Finite
Element packages Abaqus Standard, Explicit and CAE. Special tribute should be paid to Dipl.-Ing.
Michael Brunbauer, an employee of the Linz Center of Competence in Mechatronics (LCM), for some
valuable contributions.

References
[1] Abaqus Standard, Explicit, CAE / V6.3, Hibbitt, Karlsson & Sorensen, Inc., 1080 Main Street, Pawtucket,
Rhode Island, USA.
[2] N.A. Fleck, K.L. Johnson, Towards a New Theory of Cold Rolling Thin Foil, Int. J. Mech. Sci., Vol. 29,
No. 7, pp. 507-524, 1987.
[3]

S.A. Domanti, W.J. Edwards, P.J. Thomas, A Model for Foil and Thin Strip Rolling, AISE Annual
Convention, Cleveland, USA, 1994.

[4] F. Rechberger, Dressieren als Kombinierter Prge- und Walzvorgang, Doctoral Thesis at the Johannes
Kepler University Linz, Faculty of Mechatronics, 2001.
[5] W. Meindl, Walzenabplattung unter Bercksichtigung der Kontaktschubspannungen, Doctoral Thesis at the
Johannes Kepler University Linz, Faculty of Mechatronics, 2001.
[6] D. Jortner, J.F. Osterle, C.F. Zorowski, An Analysis of Cold Strip Rolling, Int. Journal of Mechanical
Sciences, Vol.2, pp. 179-194, 1960.
[7] K. Krimpelsttter, K. Zeman, G. Finstermann et al, New Advances in Temper and Skin-Pass Rolling
Technology, in: Proceedings of the 3rd European Rolling Conference (METEC Conference 03), Dsseldorf,
16-20 June, 2003.
[8] W.Y.D. Yuen, D.N. Nguyen, D.L. Matthews, Mathematical Modelling of the Temper Rolling Processes,
37th Steel Processing Conference, Ontario 1995.
[9] W.Y.D. Yuen, A. Dixon, D.N. Nguyen, The Modelling of the Mechanics of Deformation in Flat Rolling,
Journal of Materials Processing Technology, Vol. 60, pp. 87-94, 1996.
[10] H.R. Le, M.P.F. Sutcliffe, A Robust Model for Rolling of Thin Strip and Foil, Int. Journal of Mechanical
Sciences, Vol. 43, pp. 1405-1419, 2001.

8/8